1 // SPDX-License-Identifier: GPL-2.0-only
2 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
3 * Patrick Schaaf <bof@bof.de>
4 * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@netfilter.org>
7 /* Kernel module for IP set management */
9 #include <linux/init.h>
10 #include <linux/module.h>
11 #include <linux/moduleparam.h>
13 #include <linux/skbuff.h>
14 #include <linux/spinlock.h>
15 #include <linux/rculist.h>
16 #include <net/netlink.h>
17 #include <net/net_namespace.h>
18 #include <net/netns/generic.h>
20 #include <linux/netfilter.h>
21 #include <linux/netfilter/x_tables.h>
22 #include <linux/netfilter/nfnetlink.h>
23 #include <linux/netfilter/ipset/ip_set.h>
25 static LIST_HEAD(ip_set_type_list); /* all registered set types */
26 static DEFINE_MUTEX(ip_set_type_mutex); /* protects ip_set_type_list */
27 static DEFINE_RWLOCK(ip_set_ref_lock); /* protects the set refs */
30 struct ip_set * __rcu *ip_set_list; /* all individual sets */
31 ip_set_id_t ip_set_max; /* max number of sets */
32 bool is_deleted; /* deleted by ip_set_net_exit */
33 bool is_destroyed; /* all sets are destroyed */
36 static unsigned int ip_set_net_id __read_mostly;
38 static struct ip_set_net *ip_set_pernet(struct net *net)
40 return net_generic(net, ip_set_net_id);
44 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
46 static unsigned int max_sets;
48 module_param(max_sets, int, 0600);
49 MODULE_PARM_DESC(max_sets, "maximal number of sets");
50 MODULE_LICENSE("GPL");
51 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@netfilter.org>");
52 MODULE_DESCRIPTION("core IP set support");
53 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET);
55 /* When the nfnl mutex or ip_set_ref_lock is held: */
56 #define ip_set_dereference(p) \
57 rcu_dereference_protected(p, \
58 lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET) || \
59 lockdep_is_held(&ip_set_ref_lock))
60 #define ip_set(inst, id) \
61 ip_set_dereference((inst)->ip_set_list)[id]
62 #define ip_set_ref_netlink(inst,id) \
63 rcu_dereference_raw((inst)->ip_set_list)[id]
64 #define ip_set_dereference_nfnl(p) \
65 rcu_dereference_check(p, lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET))
67 /* The set types are implemented in modules and registered set types
68 * can be found in ip_set_type_list. Adding/deleting types is
69 * serialized by ip_set_type_mutex.
73 ip_set_type_lock(void)
75 mutex_lock(&ip_set_type_mutex);
79 ip_set_type_unlock(void)
81 mutex_unlock(&ip_set_type_mutex);
84 /* Register and deregister settype */
86 static struct ip_set_type *
87 find_set_type(const char *name, u8 family, u8 revision)
89 struct ip_set_type *type;
91 list_for_each_entry_rcu(type, &ip_set_type_list, list,
92 lockdep_is_held(&ip_set_type_mutex))
93 if (STRNCMP(type->name, name) &&
94 (type->family == family ||
95 type->family == NFPROTO_UNSPEC) &&
96 revision >= type->revision_min &&
97 revision <= type->revision_max)
102 /* Unlock, try to load a set type module and lock again */
104 load_settype(const char *name)
106 nfnl_unlock(NFNL_SUBSYS_IPSET);
107 pr_debug("try to load ip_set_%s\n", name);
108 if (request_module("ip_set_%s", name) < 0) {
109 pr_warn("Can't find ip_set type %s\n", name);
110 nfnl_lock(NFNL_SUBSYS_IPSET);
113 nfnl_lock(NFNL_SUBSYS_IPSET);
117 /* Find a set type and reference it */
118 #define find_set_type_get(name, family, revision, found) \
119 __find_set_type_get(name, family, revision, found, false)
122 __find_set_type_get(const char *name, u8 family, u8 revision,
123 struct ip_set_type **found, bool retry)
125 struct ip_set_type *type;
128 if (retry && !load_settype(name))
129 return -IPSET_ERR_FIND_TYPE;
132 *found = find_set_type(name, family, revision);
134 err = !try_module_get((*found)->me) ? -EFAULT : 0;
137 /* Make sure the type is already loaded
138 * but we don't support the revision
140 list_for_each_entry_rcu(type, &ip_set_type_list, list)
141 if (STRNCMP(type->name, name)) {
142 err = -IPSET_ERR_FIND_TYPE;
147 return retry ? -IPSET_ERR_FIND_TYPE :
148 __find_set_type_get(name, family, revision, found, true);
155 /* Find a given set type by name and family.
156 * If we succeeded, the supported minimal and maximum revisions are
159 #define find_set_type_minmax(name, family, min, max) \
160 __find_set_type_minmax(name, family, min, max, false)
163 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
166 struct ip_set_type *type;
169 if (retry && !load_settype(name))
170 return -IPSET_ERR_FIND_TYPE;
172 *min = 255; *max = 0;
174 list_for_each_entry_rcu(type, &ip_set_type_list, list)
175 if (STRNCMP(type->name, name) &&
176 (type->family == family ||
177 type->family == NFPROTO_UNSPEC)) {
179 if (type->revision_min < *min)
180 *min = type->revision_min;
181 if (type->revision_max > *max)
182 *max = type->revision_max;
188 return retry ? -IPSET_ERR_FIND_TYPE :
189 __find_set_type_minmax(name, family, min, max, true);
192 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
193 (f) == NFPROTO_IPV6 ? "inet6" : "any")
195 /* Register a set type structure. The type is identified by
196 * the unique triple of name, family and revision.
199 ip_set_type_register(struct ip_set_type *type)
203 if (type->protocol != IPSET_PROTOCOL) {
204 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
205 type->name, family_name(type->family),
206 type->revision_min, type->revision_max,
207 type->protocol, IPSET_PROTOCOL);
212 if (find_set_type(type->name, type->family, type->revision_min)) {
214 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
215 type->name, family_name(type->family),
217 ip_set_type_unlock();
220 list_add_rcu(&type->list, &ip_set_type_list);
221 pr_debug("type %s, family %s, revision %u:%u registered.\n",
222 type->name, family_name(type->family),
223 type->revision_min, type->revision_max);
224 ip_set_type_unlock();
228 EXPORT_SYMBOL_GPL(ip_set_type_register);
230 /* Unregister a set type. There's a small race with ip_set_create */
232 ip_set_type_unregister(struct ip_set_type *type)
235 if (!find_set_type(type->name, type->family, type->revision_min)) {
236 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
237 type->name, family_name(type->family),
239 ip_set_type_unlock();
242 list_del_rcu(&type->list);
243 pr_debug("type %s, family %s with revision min %u unregistered.\n",
244 type->name, family_name(type->family), type->revision_min);
245 ip_set_type_unlock();
249 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
251 /* Utility functions */
253 ip_set_alloc(size_t size)
255 return kvzalloc(size, GFP_KERNEL_ACCOUNT);
257 EXPORT_SYMBOL_GPL(ip_set_alloc);
260 ip_set_free(void *members)
262 pr_debug("%p: free with %s\n", members,
263 is_vmalloc_addr(members) ? "vfree" : "kfree");
266 EXPORT_SYMBOL_GPL(ip_set_free);
269 flag_nested(const struct nlattr *nla)
271 return nla->nla_type & NLA_F_NESTED;
274 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
275 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
276 [IPSET_ATTR_IPADDR_IPV6] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
280 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
282 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
284 if (unlikely(!flag_nested(nla)))
285 return -IPSET_ERR_PROTOCOL;
286 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
287 ipaddr_policy, NULL))
288 return -IPSET_ERR_PROTOCOL;
289 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
290 return -IPSET_ERR_PROTOCOL;
292 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
295 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
298 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
300 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
302 if (unlikely(!flag_nested(nla)))
303 return -IPSET_ERR_PROTOCOL;
305 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
306 ipaddr_policy, NULL))
307 return -IPSET_ERR_PROTOCOL;
308 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
309 return -IPSET_ERR_PROTOCOL;
311 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
312 sizeof(struct in6_addr));
315 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
318 ip_set_timeout_get(const unsigned long *timeout)
322 if (*timeout == IPSET_ELEM_PERMANENT)
325 t = jiffies_to_msecs(*timeout - jiffies) / MSEC_PER_SEC;
326 /* Zero value in userspace means no timeout */
327 return t == 0 ? 1 : t;
331 ip_set_comment_uget(struct nlattr *tb)
336 /* Called from uadd only, protected by the set spinlock.
337 * The kadt functions don't use the comment extensions in any way.
340 ip_set_init_comment(struct ip_set *set, struct ip_set_comment *comment,
341 const struct ip_set_ext *ext)
343 struct ip_set_comment_rcu *c = rcu_dereference_protected(comment->c, 1);
344 size_t len = ext->comment ? strlen(ext->comment) : 0;
347 set->ext_size -= sizeof(*c) + strlen(c->str) + 1;
349 rcu_assign_pointer(comment->c, NULL);
353 if (unlikely(len > IPSET_MAX_COMMENT_SIZE))
354 len = IPSET_MAX_COMMENT_SIZE;
355 c = kmalloc(sizeof(*c) + len + 1, GFP_ATOMIC);
358 strlcpy(c->str, ext->comment, len + 1);
359 set->ext_size += sizeof(*c) + strlen(c->str) + 1;
360 rcu_assign_pointer(comment->c, c);
362 EXPORT_SYMBOL_GPL(ip_set_init_comment);
364 /* Used only when dumping a set, protected by rcu_read_lock() */
366 ip_set_put_comment(struct sk_buff *skb, const struct ip_set_comment *comment)
368 struct ip_set_comment_rcu *c = rcu_dereference(comment->c);
372 return nla_put_string(skb, IPSET_ATTR_COMMENT, c->str);
375 /* Called from uadd/udel, flush or the garbage collectors protected
376 * by the set spinlock.
377 * Called when the set is destroyed and when there can't be any user
378 * of the set data anymore.
381 ip_set_comment_free(struct ip_set *set, void *ptr)
383 struct ip_set_comment *comment = ptr;
384 struct ip_set_comment_rcu *c;
386 c = rcu_dereference_protected(comment->c, 1);
389 set->ext_size -= sizeof(*c) + strlen(c->str) + 1;
391 rcu_assign_pointer(comment->c, NULL);
394 typedef void (*destroyer)(struct ip_set *, void *);
395 /* ipset data extension types, in size order */
397 const struct ip_set_ext_type ip_set_extensions[] = {
398 [IPSET_EXT_ID_COUNTER] = {
399 .type = IPSET_EXT_COUNTER,
400 .flag = IPSET_FLAG_WITH_COUNTERS,
401 .len = sizeof(struct ip_set_counter),
402 .align = __alignof__(struct ip_set_counter),
404 [IPSET_EXT_ID_TIMEOUT] = {
405 .type = IPSET_EXT_TIMEOUT,
406 .len = sizeof(unsigned long),
407 .align = __alignof__(unsigned long),
409 [IPSET_EXT_ID_SKBINFO] = {
410 .type = IPSET_EXT_SKBINFO,
411 .flag = IPSET_FLAG_WITH_SKBINFO,
412 .len = sizeof(struct ip_set_skbinfo),
413 .align = __alignof__(struct ip_set_skbinfo),
415 [IPSET_EXT_ID_COMMENT] = {
416 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
417 .flag = IPSET_FLAG_WITH_COMMENT,
418 .len = sizeof(struct ip_set_comment),
419 .align = __alignof__(struct ip_set_comment),
420 .destroy = ip_set_comment_free,
423 EXPORT_SYMBOL_GPL(ip_set_extensions);
426 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
428 return ip_set_extensions[id].flag ?
429 (flags & ip_set_extensions[id].flag) :
430 !!tb[IPSET_ATTR_TIMEOUT];
434 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
437 enum ip_set_ext_id id;
440 if (tb[IPSET_ATTR_CADT_FLAGS])
441 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
442 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
443 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
446 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
447 if (!add_extension(id, cadt_flags, tb))
449 if (align < ip_set_extensions[id].align)
450 align = ip_set_extensions[id].align;
451 len = ALIGN(len, ip_set_extensions[id].align);
452 set->offset[id] = len;
453 set->extensions |= ip_set_extensions[id].type;
454 len += ip_set_extensions[id].len;
456 return ALIGN(len, align);
458 EXPORT_SYMBOL_GPL(ip_set_elem_len);
461 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
462 struct ip_set_ext *ext)
466 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
467 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
468 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
469 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
470 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
471 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
472 return -IPSET_ERR_PROTOCOL;
474 if (tb[IPSET_ATTR_TIMEOUT]) {
475 if (!SET_WITH_TIMEOUT(set))
476 return -IPSET_ERR_TIMEOUT;
477 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
479 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
480 if (!SET_WITH_COUNTER(set))
481 return -IPSET_ERR_COUNTER;
482 if (tb[IPSET_ATTR_BYTES])
483 ext->bytes = be64_to_cpu(nla_get_be64(
484 tb[IPSET_ATTR_BYTES]));
485 if (tb[IPSET_ATTR_PACKETS])
486 ext->packets = be64_to_cpu(nla_get_be64(
487 tb[IPSET_ATTR_PACKETS]));
489 if (tb[IPSET_ATTR_COMMENT]) {
490 if (!SET_WITH_COMMENT(set))
491 return -IPSET_ERR_COMMENT;
492 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
494 if (tb[IPSET_ATTR_SKBMARK]) {
495 if (!SET_WITH_SKBINFO(set))
496 return -IPSET_ERR_SKBINFO;
497 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
498 ext->skbinfo.skbmark = fullmark >> 32;
499 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff;
501 if (tb[IPSET_ATTR_SKBPRIO]) {
502 if (!SET_WITH_SKBINFO(set))
503 return -IPSET_ERR_SKBINFO;
504 ext->skbinfo.skbprio =
505 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO]));
507 if (tb[IPSET_ATTR_SKBQUEUE]) {
508 if (!SET_WITH_SKBINFO(set))
509 return -IPSET_ERR_SKBINFO;
510 ext->skbinfo.skbqueue =
511 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE]));
515 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
518 ip_set_get_bytes(const struct ip_set_counter *counter)
520 return (u64)atomic64_read(&(counter)->bytes);
524 ip_set_get_packets(const struct ip_set_counter *counter)
526 return (u64)atomic64_read(&(counter)->packets);
530 ip_set_put_counter(struct sk_buff *skb, const struct ip_set_counter *counter)
532 return nla_put_net64(skb, IPSET_ATTR_BYTES,
533 cpu_to_be64(ip_set_get_bytes(counter)),
535 nla_put_net64(skb, IPSET_ATTR_PACKETS,
536 cpu_to_be64(ip_set_get_packets(counter)),
541 ip_set_put_skbinfo(struct sk_buff *skb, const struct ip_set_skbinfo *skbinfo)
543 /* Send nonzero parameters only */
544 return ((skbinfo->skbmark || skbinfo->skbmarkmask) &&
545 nla_put_net64(skb, IPSET_ATTR_SKBMARK,
546 cpu_to_be64((u64)skbinfo->skbmark << 32 |
547 skbinfo->skbmarkmask),
550 nla_put_net32(skb, IPSET_ATTR_SKBPRIO,
551 cpu_to_be32(skbinfo->skbprio))) ||
552 (skbinfo->skbqueue &&
553 nla_put_net16(skb, IPSET_ATTR_SKBQUEUE,
554 cpu_to_be16(skbinfo->skbqueue)));
558 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
559 const void *e, bool active)
561 if (SET_WITH_TIMEOUT(set)) {
562 unsigned long *timeout = ext_timeout(e, set);
564 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
565 htonl(active ? ip_set_timeout_get(timeout)
569 if (SET_WITH_COUNTER(set) &&
570 ip_set_put_counter(skb, ext_counter(e, set)))
572 if (SET_WITH_COMMENT(set) &&
573 ip_set_put_comment(skb, ext_comment(e, set)))
575 if (SET_WITH_SKBINFO(set) &&
576 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
580 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
583 ip_set_match_counter(u64 counter, u64 match, u8 op)
586 case IPSET_COUNTER_NONE:
588 case IPSET_COUNTER_EQ:
589 return counter == match;
590 case IPSET_COUNTER_NE:
591 return counter != match;
592 case IPSET_COUNTER_LT:
593 return counter < match;
594 case IPSET_COUNTER_GT:
595 return counter > match;
601 ip_set_add_bytes(u64 bytes, struct ip_set_counter *counter)
603 atomic64_add((long long)bytes, &(counter)->bytes);
607 ip_set_add_packets(u64 packets, struct ip_set_counter *counter)
609 atomic64_add((long long)packets, &(counter)->packets);
613 ip_set_update_counter(struct ip_set_counter *counter,
614 const struct ip_set_ext *ext, u32 flags)
616 if (ext->packets != ULLONG_MAX &&
617 !(flags & IPSET_FLAG_SKIP_COUNTER_UPDATE)) {
618 ip_set_add_bytes(ext->bytes, counter);
619 ip_set_add_packets(ext->packets, counter);
624 ip_set_get_skbinfo(struct ip_set_skbinfo *skbinfo,
625 const struct ip_set_ext *ext,
626 struct ip_set_ext *mext, u32 flags)
628 mext->skbinfo = *skbinfo;
632 ip_set_match_extensions(struct ip_set *set, const struct ip_set_ext *ext,
633 struct ip_set_ext *mext, u32 flags, void *data)
635 if (SET_WITH_TIMEOUT(set) &&
636 ip_set_timeout_expired(ext_timeout(data, set)))
638 if (SET_WITH_COUNTER(set)) {
639 struct ip_set_counter *counter = ext_counter(data, set);
641 ip_set_update_counter(counter, ext, flags);
643 if (flags & IPSET_FLAG_MATCH_COUNTERS &&
644 !(ip_set_match_counter(ip_set_get_packets(counter),
645 mext->packets, mext->packets_op) &&
646 ip_set_match_counter(ip_set_get_bytes(counter),
647 mext->bytes, mext->bytes_op)))
650 if (SET_WITH_SKBINFO(set))
651 ip_set_get_skbinfo(ext_skbinfo(data, set),
655 EXPORT_SYMBOL_GPL(ip_set_match_extensions);
657 /* Creating/destroying/renaming/swapping affect the existence and
658 * the properties of a set. All of these can be executed from userspace
659 * only and serialized by the nfnl mutex indirectly from nfnetlink.
661 * Sets are identified by their index in ip_set_list and the index
662 * is used by the external references (set/SET netfilter modules).
664 * The set behind an index may change by swapping only, from userspace.
668 __ip_set_get(struct ip_set *set)
670 write_lock_bh(&ip_set_ref_lock);
672 write_unlock_bh(&ip_set_ref_lock);
676 __ip_set_put(struct ip_set *set)
678 write_lock_bh(&ip_set_ref_lock);
679 BUG_ON(set->ref == 0);
681 write_unlock_bh(&ip_set_ref_lock);
684 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
685 * a separate reference counter
688 __ip_set_get_netlink(struct ip_set *set)
690 write_lock_bh(&ip_set_ref_lock);
692 write_unlock_bh(&ip_set_ref_lock);
696 __ip_set_put_netlink(struct ip_set *set)
698 write_lock_bh(&ip_set_ref_lock);
699 BUG_ON(set->ref_netlink == 0);
701 write_unlock_bh(&ip_set_ref_lock);
704 /* Add, del and test set entries from kernel.
706 * The set behind the index must exist and must be referenced
707 * so it can't be destroyed (or changed) under our foot.
710 static struct ip_set *
711 ip_set_rcu_get(struct net *net, ip_set_id_t index)
713 struct ip_set_net *inst = ip_set_pernet(net);
715 /* ip_set_list and the set pointer need to be protected */
716 return ip_set_dereference_nfnl(inst->ip_set_list)[index];
720 ip_set_lock(struct ip_set *set)
722 if (!set->variant->region_lock)
723 spin_lock_bh(&set->lock);
727 ip_set_unlock(struct ip_set *set)
729 if (!set->variant->region_lock)
730 spin_unlock_bh(&set->lock);
734 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
735 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
737 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
741 pr_debug("set %s, index %u\n", set->name, index);
743 if (opt->dim < set->type->dimension ||
744 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
748 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
749 rcu_read_unlock_bh();
751 if (ret == -EAGAIN) {
752 /* Type requests element to be completed */
753 pr_debug("element must be completed, ADD is triggered\n");
755 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
759 /* --return-nomatch: invert matched element */
760 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
761 (set->type->features & IPSET_TYPE_NOMATCH) &&
762 (ret > 0 || ret == -ENOTEMPTY))
766 /* Convert error codes to nomatch */
767 return (ret < 0 ? 0 : ret);
769 EXPORT_SYMBOL_GPL(ip_set_test);
772 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
773 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
775 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
779 pr_debug("set %s, index %u\n", set->name, index);
781 if (opt->dim < set->type->dimension ||
782 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
783 return -IPSET_ERR_TYPE_MISMATCH;
786 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
791 EXPORT_SYMBOL_GPL(ip_set_add);
794 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
795 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
797 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
801 pr_debug("set %s, index %u\n", set->name, index);
803 if (opt->dim < set->type->dimension ||
804 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
805 return -IPSET_ERR_TYPE_MISMATCH;
808 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
813 EXPORT_SYMBOL_GPL(ip_set_del);
815 /* Find set by name, reference it once. The reference makes sure the
816 * thing pointed to, does not go away under our feet.
820 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
822 ip_set_id_t i, index = IPSET_INVALID_ID;
824 struct ip_set_net *inst = ip_set_pernet(net);
827 for (i = 0; i < inst->ip_set_max; i++) {
828 s = rcu_dereference(inst->ip_set_list)[i];
829 if (s && STRNCMP(s->name, name)) {
840 EXPORT_SYMBOL_GPL(ip_set_get_byname);
842 /* If the given set pointer points to a valid set, decrement
843 * reference count by 1. The caller shall not assume the index
844 * to be valid, after calling this function.
849 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
854 set = rcu_dereference(inst->ip_set_list)[index];
861 ip_set_put_byindex(struct net *net, ip_set_id_t index)
863 struct ip_set_net *inst = ip_set_pernet(net);
865 __ip_set_put_byindex(inst, index);
867 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
869 /* Get the name of a set behind a set index.
870 * Set itself is protected by RCU, but its name isn't: to protect against
871 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
875 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name)
877 struct ip_set *set = ip_set_rcu_get(net, index);
881 read_lock_bh(&ip_set_ref_lock);
882 strncpy(name, set->name, IPSET_MAXNAMELEN);
883 read_unlock_bh(&ip_set_ref_lock);
885 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
887 /* Routines to call by external subsystems, which do not
888 * call nfnl_lock for us.
891 /* Find set by index, reference it once. The reference makes sure the
892 * thing pointed to, does not go away under our feet.
894 * The nfnl mutex is used in the function.
897 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
900 struct ip_set_net *inst = ip_set_pernet(net);
902 if (index >= inst->ip_set_max)
903 return IPSET_INVALID_ID;
905 nfnl_lock(NFNL_SUBSYS_IPSET);
906 set = ip_set(inst, index);
910 index = IPSET_INVALID_ID;
911 nfnl_unlock(NFNL_SUBSYS_IPSET);
915 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
917 /* If the given set pointer points to a valid set, decrement
918 * reference count by 1. The caller shall not assume the index
919 * to be valid, after calling this function.
921 * The nfnl mutex is used in the function.
924 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
927 struct ip_set_net *inst = ip_set_pernet(net);
929 nfnl_lock(NFNL_SUBSYS_IPSET);
930 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
931 set = ip_set(inst, index);
935 nfnl_unlock(NFNL_SUBSYS_IPSET);
937 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
939 /* Communication protocol with userspace over netlink.
941 * The commands are serialized by the nfnl mutex.
944 static inline u8 protocol(const struct nlattr * const tb[])
946 return nla_get_u8(tb[IPSET_ATTR_PROTOCOL]);
950 protocol_failed(const struct nlattr * const tb[])
952 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) != IPSET_PROTOCOL;
956 protocol_min_failed(const struct nlattr * const tb[])
958 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) < IPSET_PROTOCOL_MIN;
962 flag_exist(const struct nlmsghdr *nlh)
964 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
967 static struct nlmsghdr *
968 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
971 return nfnl_msg_put(skb, portid, seq,
972 nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd), flags,
973 NFPROTO_IPV4, NFNETLINK_V0, 0);
978 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
979 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
980 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
981 .len = IPSET_MAXNAMELEN - 1 },
982 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
983 .len = IPSET_MAXNAMELEN - 1},
984 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
985 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
986 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
989 static struct ip_set *
990 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
992 struct ip_set *set = NULL;
995 *id = IPSET_INVALID_ID;
996 for (i = 0; i < inst->ip_set_max; i++) {
997 set = ip_set(inst, i);
998 if (set && STRNCMP(set->name, name)) {
1003 return (*id == IPSET_INVALID_ID ? NULL : set);
1006 static inline struct ip_set *
1007 find_set(struct ip_set_net *inst, const char *name)
1011 return find_set_and_id(inst, name, &id);
1015 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
1016 struct ip_set **set)
1021 *index = IPSET_INVALID_ID;
1022 for (i = 0; i < inst->ip_set_max; i++) {
1023 s = ip_set(inst, i);
1025 if (*index == IPSET_INVALID_ID)
1027 } else if (STRNCMP(name, s->name)) {
1033 if (*index == IPSET_INVALID_ID)
1034 /* No free slot remained */
1035 return -IPSET_ERR_MAX_SETS;
1039 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1040 const struct nlmsghdr *nlh,
1041 const struct nlattr * const attr[],
1042 struct netlink_ext_ack *extack)
1047 static int ip_set_create(struct net *net, struct sock *ctnl,
1048 struct sk_buff *skb, const struct nlmsghdr *nlh,
1049 const struct nlattr * const attr[],
1050 struct netlink_ext_ack *extack)
1052 struct ip_set_net *inst = ip_set_pernet(net);
1053 struct ip_set *set, *clash = NULL;
1054 ip_set_id_t index = IPSET_INVALID_ID;
1055 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
1056 const char *name, *typename;
1057 u8 family, revision;
1058 u32 flags = flag_exist(nlh);
1061 if (unlikely(protocol_min_failed(attr) ||
1062 !attr[IPSET_ATTR_SETNAME] ||
1063 !attr[IPSET_ATTR_TYPENAME] ||
1064 !attr[IPSET_ATTR_REVISION] ||
1065 !attr[IPSET_ATTR_FAMILY] ||
1066 (attr[IPSET_ATTR_DATA] &&
1067 !flag_nested(attr[IPSET_ATTR_DATA]))))
1068 return -IPSET_ERR_PROTOCOL;
1070 name = nla_data(attr[IPSET_ATTR_SETNAME]);
1071 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1072 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1073 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
1074 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
1075 name, typename, family_name(family), revision);
1077 /* First, and without any locks, allocate and initialize
1078 * a normal base set structure.
1080 set = kzalloc(sizeof(*set), GFP_KERNEL);
1083 spin_lock_init(&set->lock);
1084 strlcpy(set->name, name, IPSET_MAXNAMELEN);
1085 set->family = family;
1086 set->revision = revision;
1088 /* Next, check that we know the type, and take
1089 * a reference on the type, to make sure it stays available
1090 * while constructing our new set.
1092 * After referencing the type, we try to create the type
1093 * specific part of the set without holding any locks.
1095 ret = find_set_type_get(typename, family, revision, &set->type);
1099 /* Without holding any locks, create private part. */
1100 if (attr[IPSET_ATTR_DATA] &&
1101 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
1102 set->type->create_policy, NULL)) {
1103 ret = -IPSET_ERR_PROTOCOL;
1107 ret = set->type->create(net, set, tb, flags);
1111 /* BTW, ret==0 here. */
1113 /* Here, we have a valid, constructed set and we are protected
1114 * by the nfnl mutex. Find the first free index in ip_set_list
1115 * and check clashing.
1117 ret = find_free_id(inst, set->name, &index, &clash);
1118 if (ret == -EEXIST) {
1119 /* If this is the same set and requested, ignore error */
1120 if ((flags & IPSET_FLAG_EXIST) &&
1121 STRNCMP(set->type->name, clash->type->name) &&
1122 set->type->family == clash->type->family &&
1123 set->type->revision_min == clash->type->revision_min &&
1124 set->type->revision_max == clash->type->revision_max &&
1125 set->variant->same_set(set, clash))
1128 } else if (ret == -IPSET_ERR_MAX_SETS) {
1129 struct ip_set **list, **tmp;
1130 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
1132 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
1136 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
1139 /* nfnl mutex is held, both lists are valid */
1140 tmp = ip_set_dereference(inst->ip_set_list);
1141 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
1142 rcu_assign_pointer(inst->ip_set_list, list);
1143 /* Make sure all current packets have passed through */
1146 index = inst->ip_set_max;
1147 inst->ip_set_max = i;
1154 /* Finally! Add our shiny new set to the list, and be done. */
1155 pr_debug("create: '%s' created with index %u!\n", set->name, index);
1156 ip_set(inst, index) = set;
1161 set->variant->cancel_gc(set);
1162 set->variant->destroy(set);
1164 module_put(set->type->me);
1172 static const struct nla_policy
1173 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
1174 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1175 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1176 .len = IPSET_MAXNAMELEN - 1 },
1180 ip_set_destroy_set(struct ip_set *set)
1182 pr_debug("set: %s\n", set->name);
1184 /* Must call it without holding any lock */
1185 set->variant->destroy(set);
1186 module_put(set->type->me);
1191 ip_set_destroy_set_rcu(struct rcu_head *head)
1193 struct ip_set *set = container_of(head, struct ip_set, rcu);
1195 ip_set_destroy_set(set);
1198 static int ip_set_destroy(struct net *net, struct sock *ctnl,
1199 struct sk_buff *skb, const struct nlmsghdr *nlh,
1200 const struct nlattr * const attr[],
1201 struct netlink_ext_ack *extack)
1203 struct ip_set_net *inst = ip_set_pernet(net);
1208 if (unlikely(protocol_min_failed(attr)))
1209 return -IPSET_ERR_PROTOCOL;
1212 /* Commands are serialized and references are
1213 * protected by the ip_set_ref_lock.
1214 * External systems (i.e. xt_set) must call
1215 * ip_set_put|get_nfnl_* functions, that way we
1216 * can safely check references here.
1218 * list:set timer can only decrement the reference
1219 * counter, so if it's already zero, we can proceed
1220 * without holding the lock.
1222 if (!attr[IPSET_ATTR_SETNAME]) {
1223 /* Must wait for flush to be really finished in list:set */
1225 read_lock_bh(&ip_set_ref_lock);
1226 for (i = 0; i < inst->ip_set_max; i++) {
1227 s = ip_set(inst, i);
1228 if (s && (s->ref || s->ref_netlink)) {
1229 ret = -IPSET_ERR_BUSY;
1233 inst->is_destroyed = true;
1234 read_unlock_bh(&ip_set_ref_lock);
1235 for (i = 0; i < inst->ip_set_max; i++) {
1236 s = ip_set(inst, i);
1238 ip_set(inst, i) = NULL;
1239 /* Must cancel garbage collectors */
1240 s->variant->cancel_gc(s);
1241 ip_set_destroy_set(s);
1244 /* Modified by ip_set_destroy() only, which is serialized */
1245 inst->is_destroyed = false;
1249 read_lock_bh(&ip_set_ref_lock);
1250 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1255 } else if (s->ref || s->ref_netlink) {
1256 ret = -IPSET_ERR_BUSY;
1259 features = s->type->features;
1260 ip_set(inst, i) = NULL;
1261 read_unlock_bh(&ip_set_ref_lock);
1262 if (features & IPSET_TYPE_NAME) {
1263 /* Must wait for flush to be really finished */
1266 /* Must cancel garbage collectors */
1267 s->variant->cancel_gc(s);
1268 call_rcu(&s->rcu, ip_set_destroy_set_rcu);
1272 read_unlock_bh(&ip_set_ref_lock);
1279 ip_set_flush_set(struct ip_set *set)
1281 pr_debug("set: %s\n", set->name);
1284 set->variant->flush(set);
1288 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1289 const struct nlmsghdr *nlh,
1290 const struct nlattr * const attr[],
1291 struct netlink_ext_ack *extack)
1293 struct ip_set_net *inst = ip_set_pernet(net);
1297 if (unlikely(protocol_min_failed(attr)))
1298 return -IPSET_ERR_PROTOCOL;
1300 if (!attr[IPSET_ATTR_SETNAME]) {
1301 for (i = 0; i < inst->ip_set_max; i++) {
1302 s = ip_set(inst, i);
1304 ip_set_flush_set(s);
1307 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1311 ip_set_flush_set(s);
1319 static const struct nla_policy
1320 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1321 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1322 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1323 .len = IPSET_MAXNAMELEN - 1 },
1324 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1325 .len = IPSET_MAXNAMELEN - 1 },
1328 static int ip_set_rename(struct net *net, struct sock *ctnl,
1329 struct sk_buff *skb, const struct nlmsghdr *nlh,
1330 const struct nlattr * const attr[],
1331 struct netlink_ext_ack *extack)
1333 struct ip_set_net *inst = ip_set_pernet(net);
1334 struct ip_set *set, *s;
1339 if (unlikely(protocol_min_failed(attr) ||
1340 !attr[IPSET_ATTR_SETNAME] ||
1341 !attr[IPSET_ATTR_SETNAME2]))
1342 return -IPSET_ERR_PROTOCOL;
1344 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1348 write_lock_bh(&ip_set_ref_lock);
1349 if (set->ref != 0 || set->ref_netlink != 0) {
1350 ret = -IPSET_ERR_REFERENCED;
1354 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1355 for (i = 0; i < inst->ip_set_max; i++) {
1356 s = ip_set(inst, i);
1357 if (s && STRNCMP(s->name, name2)) {
1358 ret = -IPSET_ERR_EXIST_SETNAME2;
1362 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1365 write_unlock_bh(&ip_set_ref_lock);
1369 /* Swap two sets so that name/index points to the other.
1370 * References and set names are also swapped.
1372 * The commands are serialized by the nfnl mutex and references are
1373 * protected by the ip_set_ref_lock. The kernel interfaces
1374 * do not hold the mutex but the pointer settings are atomic
1375 * so the ip_set_list always contains valid pointers to the sets.
1378 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1379 const struct nlmsghdr *nlh,
1380 const struct nlattr * const attr[],
1381 struct netlink_ext_ack *extack)
1383 struct ip_set_net *inst = ip_set_pernet(net);
1384 struct ip_set *from, *to;
1385 ip_set_id_t from_id, to_id;
1386 char from_name[IPSET_MAXNAMELEN];
1388 if (unlikely(protocol_min_failed(attr) ||
1389 !attr[IPSET_ATTR_SETNAME] ||
1390 !attr[IPSET_ATTR_SETNAME2]))
1391 return -IPSET_ERR_PROTOCOL;
1393 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1398 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1401 return -IPSET_ERR_EXIST_SETNAME2;
1403 /* Features must not change.
1404 * Not an artifical restriction anymore, as we must prevent
1405 * possible loops created by swapping in setlist type of sets.
1407 if (!(from->type->features == to->type->features &&
1408 from->family == to->family))
1409 return -IPSET_ERR_TYPE_MISMATCH;
1411 write_lock_bh(&ip_set_ref_lock);
1413 if (from->ref_netlink || to->ref_netlink) {
1414 write_unlock_bh(&ip_set_ref_lock);
1418 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1419 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1420 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1422 swap(from->ref, to->ref);
1423 ip_set(inst, from_id) = to;
1424 ip_set(inst, to_id) = from;
1425 write_unlock_bh(&ip_set_ref_lock);
1430 /* List/save set data */
1437 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1438 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1441 ip_set_put_flags(struct sk_buff *skb, struct ip_set *set)
1445 if (SET_WITH_TIMEOUT(set))
1446 if (unlikely(nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
1447 htonl(set->timeout))))
1449 if (SET_WITH_COUNTER(set))
1450 cadt_flags |= IPSET_FLAG_WITH_COUNTERS;
1451 if (SET_WITH_COMMENT(set))
1452 cadt_flags |= IPSET_FLAG_WITH_COMMENT;
1453 if (SET_WITH_SKBINFO(set))
1454 cadt_flags |= IPSET_FLAG_WITH_SKBINFO;
1455 if (SET_WITH_FORCEADD(set))
1456 cadt_flags |= IPSET_FLAG_WITH_FORCEADD;
1460 return nla_put_net32(skb, IPSET_ATTR_CADT_FLAGS, htonl(cadt_flags));
1462 EXPORT_SYMBOL_GPL(ip_set_put_flags);
1465 ip_set_dump_done(struct netlink_callback *cb)
1467 if (cb->args[IPSET_CB_ARG0]) {
1468 struct ip_set_net *inst =
1469 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1470 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1471 struct ip_set *set = ip_set_ref_netlink(inst, index);
1473 if (set->variant->uref)
1474 set->variant->uref(set, cb, false);
1475 pr_debug("release set %s\n", set->name);
1476 __ip_set_put_netlink(set);
1482 dump_attrs(struct nlmsghdr *nlh)
1484 const struct nlattr *attr;
1487 pr_debug("dump nlmsg\n");
1488 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1489 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1493 static const struct nla_policy
1494 ip_set_dump_policy[IPSET_ATTR_CMD_MAX + 1] = {
1495 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1496 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1497 .len = IPSET_MAXNAMELEN - 1 },
1498 [IPSET_ATTR_FLAGS] = { .type = NLA_U32 },
1502 ip_set_dump_start(struct netlink_callback *cb)
1504 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1505 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1506 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1507 struct nlattr *attr = (void *)nlh + min_len;
1508 struct sk_buff *skb = cb->skb;
1509 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1513 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, attr,
1514 nlh->nlmsg_len - min_len,
1515 ip_set_dump_policy, NULL);
1519 cb->args[IPSET_CB_PROTO] = nla_get_u8(cda[IPSET_ATTR_PROTOCOL]);
1520 if (cda[IPSET_ATTR_SETNAME]) {
1524 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1530 dump_type = DUMP_ONE;
1531 cb->args[IPSET_CB_INDEX] = index;
1533 dump_type = DUMP_ALL;
1536 if (cda[IPSET_ATTR_FLAGS]) {
1537 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1539 dump_type |= (f << 16);
1541 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1542 cb->args[IPSET_CB_DUMP] = dump_type;
1547 /* We have to create and send the error message manually :-( */
1548 if (nlh->nlmsg_flags & NLM_F_ACK) {
1549 netlink_ack(cb->skb, nlh, ret, NULL);
1555 ip_set_dump_do(struct sk_buff *skb, struct netlink_callback *cb)
1557 ip_set_id_t index = IPSET_INVALID_ID, max;
1558 struct ip_set *set = NULL;
1559 struct nlmsghdr *nlh = NULL;
1560 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1561 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1562 u32 dump_type, dump_flags;
1566 if (!cb->args[IPSET_CB_DUMP])
1569 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1572 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1573 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1574 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1577 pr_debug("dump type, flag: %u %u index: %ld\n",
1578 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1579 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1580 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1581 write_lock_bh(&ip_set_ref_lock);
1582 set = ip_set(inst, index);
1583 is_destroyed = inst->is_destroyed;
1584 if (!set || is_destroyed) {
1585 write_unlock_bh(&ip_set_ref_lock);
1586 if (dump_type == DUMP_ONE) {
1591 /* All sets are just being destroyed */
1597 /* When dumping all sets, we must dump "sorted"
1598 * so that lists (unions of sets) are dumped last.
1600 if (dump_type != DUMP_ONE &&
1601 ((dump_type == DUMP_ALL) ==
1602 !!(set->type->features & IPSET_DUMP_LAST))) {
1603 write_unlock_bh(&ip_set_ref_lock);
1606 pr_debug("List set: %s\n", set->name);
1607 if (!cb->args[IPSET_CB_ARG0]) {
1608 /* Start listing: make sure set won't be destroyed */
1609 pr_debug("reference set\n");
1612 write_unlock_bh(&ip_set_ref_lock);
1613 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1614 cb->nlh->nlmsg_seq, flags,
1618 goto release_refcount;
1620 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL,
1621 cb->args[IPSET_CB_PROTO]) ||
1622 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1623 goto nla_put_failure;
1624 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1626 switch (cb->args[IPSET_CB_ARG0]) {
1628 /* Core header data */
1629 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1631 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1633 nla_put_u8(skb, IPSET_ATTR_REVISION,
1635 goto nla_put_failure;
1636 if (cb->args[IPSET_CB_PROTO] > IPSET_PROTOCOL_MIN &&
1637 nla_put_net16(skb, IPSET_ATTR_INDEX, htons(index)))
1638 goto nla_put_failure;
1639 ret = set->variant->head(set, skb);
1641 goto release_refcount;
1642 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1644 if (set->variant->uref)
1645 set->variant->uref(set, cb, true);
1648 ret = set->variant->list(set, skb, cb);
1649 if (!cb->args[IPSET_CB_ARG0])
1650 /* Set is done, proceed with next one */
1652 goto release_refcount;
1655 /* If we dump all sets, continue with dumping last ones */
1656 if (dump_type == DUMP_ALL) {
1657 dump_type = DUMP_LAST;
1658 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1659 cb->args[IPSET_CB_INDEX] = 0;
1660 if (set && set->variant->uref)
1661 set->variant->uref(set, cb, false);
1669 if (dump_type == DUMP_ONE)
1670 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1672 cb->args[IPSET_CB_INDEX]++;
1674 /* If there was an error or set is done, release set */
1675 if (ret || !cb->args[IPSET_CB_ARG0]) {
1676 set = ip_set_ref_netlink(inst, index);
1677 if (set->variant->uref)
1678 set->variant->uref(set, cb, false);
1679 pr_debug("release set %s\n", set->name);
1680 __ip_set_put_netlink(set);
1681 cb->args[IPSET_CB_ARG0] = 0;
1685 nlmsg_end(skb, nlh);
1686 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1690 return ret < 0 ? ret : skb->len;
1693 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1694 const struct nlmsghdr *nlh,
1695 const struct nlattr * const attr[],
1696 struct netlink_ext_ack *extack)
1698 if (unlikely(protocol_min_failed(attr)))
1699 return -IPSET_ERR_PROTOCOL;
1702 struct netlink_dump_control c = {
1703 .start = ip_set_dump_start,
1704 .dump = ip_set_dump_do,
1705 .done = ip_set_dump_done,
1707 return netlink_dump_start(ctnl, skb, nlh, &c);
1711 /* Add, del and test */
1713 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1714 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1715 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1716 .len = IPSET_MAXNAMELEN - 1 },
1717 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1718 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1719 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1723 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1724 struct nlattr *tb[], enum ipset_adt adt,
1725 u32 flags, bool use_lineno)
1729 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1733 __ip_set_get_netlink(set);
1734 nfnl_unlock(NFNL_SUBSYS_IPSET);
1736 nfnl_lock(NFNL_SUBSYS_IPSET);
1737 __ip_set_put_netlink(set);
1741 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1744 } while (ret == -ERANGE ||
1746 set->variant->resize &&
1747 (ret = set->variant->resize(set, retried)) == 0));
1749 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1751 if (lineno && use_lineno) {
1752 /* Error in restore/batch mode: send back lineno */
1753 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1754 struct sk_buff *skb2;
1755 struct nlmsgerr *errmsg;
1756 size_t payload = min(SIZE_MAX,
1757 sizeof(*errmsg) + nlmsg_len(nlh));
1758 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1759 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1760 struct nlattr *cmdattr;
1763 skb2 = nlmsg_new(payload, GFP_KERNEL);
1766 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1767 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1768 errmsg = nlmsg_data(rep);
1769 errmsg->error = ret;
1770 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1771 cmdattr = (void *)&errmsg->msg + min_len;
1773 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1774 nlh->nlmsg_len - min_len, ip_set_adt_policy,
1781 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1785 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1787 /* Signal netlink not to send its ACK/errmsg. */
1794 static int ip_set_ad(struct net *net, struct sock *ctnl,
1795 struct sk_buff *skb,
1797 const struct nlmsghdr *nlh,
1798 const struct nlattr * const attr[],
1799 struct netlink_ext_ack *extack)
1801 struct ip_set_net *inst = ip_set_pernet(net);
1803 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1804 const struct nlattr *nla;
1805 u32 flags = flag_exist(nlh);
1809 if (unlikely(protocol_min_failed(attr) ||
1810 !attr[IPSET_ATTR_SETNAME] ||
1811 !((attr[IPSET_ATTR_DATA] != NULL) ^
1812 (attr[IPSET_ATTR_ADT] != NULL)) ||
1813 (attr[IPSET_ATTR_DATA] &&
1814 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1815 (attr[IPSET_ATTR_ADT] &&
1816 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1817 !attr[IPSET_ATTR_LINENO]))))
1818 return -IPSET_ERR_PROTOCOL;
1820 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1824 use_lineno = !!attr[IPSET_ATTR_LINENO];
1825 if (attr[IPSET_ATTR_DATA]) {
1826 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1827 attr[IPSET_ATTR_DATA],
1828 set->type->adt_policy, NULL))
1829 return -IPSET_ERR_PROTOCOL;
1830 ret = call_ad(ctnl, skb, set, tb, adt, flags,
1835 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1836 if (nla_type(nla) != IPSET_ATTR_DATA ||
1837 !flag_nested(nla) ||
1838 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1839 set->type->adt_policy, NULL))
1840 return -IPSET_ERR_PROTOCOL;
1841 ret = call_ad(ctnl, skb, set, tb, adt,
1850 static int ip_set_uadd(struct net *net, struct sock *ctnl,
1851 struct sk_buff *skb, const struct nlmsghdr *nlh,
1852 const struct nlattr * const attr[],
1853 struct netlink_ext_ack *extack)
1855 return ip_set_ad(net, ctnl, skb,
1856 IPSET_ADD, nlh, attr, extack);
1859 static int ip_set_udel(struct net *net, struct sock *ctnl,
1860 struct sk_buff *skb, const struct nlmsghdr *nlh,
1861 const struct nlattr * const attr[],
1862 struct netlink_ext_ack *extack)
1864 return ip_set_ad(net, ctnl, skb,
1865 IPSET_DEL, nlh, attr, extack);
1868 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1869 const struct nlmsghdr *nlh,
1870 const struct nlattr * const attr[],
1871 struct netlink_ext_ack *extack)
1873 struct ip_set_net *inst = ip_set_pernet(net);
1875 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1879 if (unlikely(protocol_min_failed(attr) ||
1880 !attr[IPSET_ATTR_SETNAME] ||
1881 !attr[IPSET_ATTR_DATA] ||
1882 !flag_nested(attr[IPSET_ATTR_DATA])))
1883 return -IPSET_ERR_PROTOCOL;
1885 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1889 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1890 set->type->adt_policy, NULL))
1891 return -IPSET_ERR_PROTOCOL;
1894 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1895 rcu_read_unlock_bh();
1896 /* Userspace can't trigger element to be re-added */
1900 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1903 /* Get headed data of a set */
1905 static int ip_set_header(struct net *net, struct sock *ctnl,
1906 struct sk_buff *skb, const struct nlmsghdr *nlh,
1907 const struct nlattr * const attr[],
1908 struct netlink_ext_ack *extack)
1910 struct ip_set_net *inst = ip_set_pernet(net);
1911 const struct ip_set *set;
1912 struct sk_buff *skb2;
1913 struct nlmsghdr *nlh2;
1916 if (unlikely(protocol_min_failed(attr) ||
1917 !attr[IPSET_ATTR_SETNAME]))
1918 return -IPSET_ERR_PROTOCOL;
1920 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1924 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1928 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1932 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1933 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1934 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1935 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1936 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1937 goto nla_put_failure;
1938 nlmsg_end(skb2, nlh2);
1940 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1947 nlmsg_cancel(skb2, nlh2);
1955 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1956 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1957 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1958 .len = IPSET_MAXNAMELEN - 1 },
1959 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1962 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1963 const struct nlmsghdr *nlh,
1964 const struct nlattr * const attr[],
1965 struct netlink_ext_ack *extack)
1967 struct sk_buff *skb2;
1968 struct nlmsghdr *nlh2;
1969 u8 family, min, max;
1970 const char *typename;
1973 if (unlikely(protocol_min_failed(attr) ||
1974 !attr[IPSET_ATTR_TYPENAME] ||
1975 !attr[IPSET_ATTR_FAMILY]))
1976 return -IPSET_ERR_PROTOCOL;
1978 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1979 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1980 ret = find_set_type_minmax(typename, family, &min, &max);
1984 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1988 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1992 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1993 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1994 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1995 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1996 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1997 goto nla_put_failure;
1998 nlmsg_end(skb2, nlh2);
2000 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
2001 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2008 nlmsg_cancel(skb2, nlh2);
2014 /* Get protocol version */
2016 static const struct nla_policy
2017 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
2018 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
2021 static int ip_set_protocol(struct net *net, struct sock *ctnl,
2022 struct sk_buff *skb, const struct nlmsghdr *nlh,
2023 const struct nlattr * const attr[],
2024 struct netlink_ext_ack *extack)
2026 struct sk_buff *skb2;
2027 struct nlmsghdr *nlh2;
2030 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
2031 return -IPSET_ERR_PROTOCOL;
2033 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2037 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2038 IPSET_CMD_PROTOCOL);
2041 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
2042 goto nla_put_failure;
2043 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL_MIN, IPSET_PROTOCOL_MIN))
2044 goto nla_put_failure;
2045 nlmsg_end(skb2, nlh2);
2047 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2054 nlmsg_cancel(skb2, nlh2);
2060 /* Get set by name or index, from userspace */
2062 static int ip_set_byname(struct net *net, struct sock *ctnl,
2063 struct sk_buff *skb, const struct nlmsghdr *nlh,
2064 const struct nlattr * const attr[],
2065 struct netlink_ext_ack *extack)
2067 struct ip_set_net *inst = ip_set_pernet(net);
2068 struct sk_buff *skb2;
2069 struct nlmsghdr *nlh2;
2070 ip_set_id_t id = IPSET_INVALID_ID;
2071 const struct ip_set *set;
2074 if (unlikely(protocol_failed(attr) ||
2075 !attr[IPSET_ATTR_SETNAME]))
2076 return -IPSET_ERR_PROTOCOL;
2078 set = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), &id);
2079 if (id == IPSET_INVALID_ID)
2082 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2086 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2087 IPSET_CMD_GET_BYNAME);
2090 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2091 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
2092 nla_put_net16(skb2, IPSET_ATTR_INDEX, htons(id)))
2093 goto nla_put_failure;
2094 nlmsg_end(skb2, nlh2);
2096 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2103 nlmsg_cancel(skb2, nlh2);
2109 static const struct nla_policy ip_set_index_policy[IPSET_ATTR_CMD_MAX + 1] = {
2110 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
2111 [IPSET_ATTR_INDEX] = { .type = NLA_U16 },
2114 static int ip_set_byindex(struct net *net, struct sock *ctnl,
2115 struct sk_buff *skb, const struct nlmsghdr *nlh,
2116 const struct nlattr * const attr[],
2117 struct netlink_ext_ack *extack)
2119 struct ip_set_net *inst = ip_set_pernet(net);
2120 struct sk_buff *skb2;
2121 struct nlmsghdr *nlh2;
2122 ip_set_id_t id = IPSET_INVALID_ID;
2123 const struct ip_set *set;
2126 if (unlikely(protocol_failed(attr) ||
2127 !attr[IPSET_ATTR_INDEX]))
2128 return -IPSET_ERR_PROTOCOL;
2130 id = ip_set_get_h16(attr[IPSET_ATTR_INDEX]);
2131 if (id >= inst->ip_set_max)
2133 set = ip_set(inst, id);
2137 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2141 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2142 IPSET_CMD_GET_BYINDEX);
2145 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2146 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name))
2147 goto nla_put_failure;
2148 nlmsg_end(skb2, nlh2);
2150 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2157 nlmsg_cancel(skb2, nlh2);
2163 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
2164 [IPSET_CMD_NONE] = {
2165 .call = ip_set_none,
2166 .attr_count = IPSET_ATTR_CMD_MAX,
2168 [IPSET_CMD_CREATE] = {
2169 .call = ip_set_create,
2170 .attr_count = IPSET_ATTR_CMD_MAX,
2171 .policy = ip_set_create_policy,
2173 [IPSET_CMD_DESTROY] = {
2174 .call = ip_set_destroy,
2175 .attr_count = IPSET_ATTR_CMD_MAX,
2176 .policy = ip_set_setname_policy,
2178 [IPSET_CMD_FLUSH] = {
2179 .call = ip_set_flush,
2180 .attr_count = IPSET_ATTR_CMD_MAX,
2181 .policy = ip_set_setname_policy,
2183 [IPSET_CMD_RENAME] = {
2184 .call = ip_set_rename,
2185 .attr_count = IPSET_ATTR_CMD_MAX,
2186 .policy = ip_set_setname2_policy,
2188 [IPSET_CMD_SWAP] = {
2189 .call = ip_set_swap,
2190 .attr_count = IPSET_ATTR_CMD_MAX,
2191 .policy = ip_set_setname2_policy,
2193 [IPSET_CMD_LIST] = {
2194 .call = ip_set_dump,
2195 .attr_count = IPSET_ATTR_CMD_MAX,
2196 .policy = ip_set_dump_policy,
2198 [IPSET_CMD_SAVE] = {
2199 .call = ip_set_dump,
2200 .attr_count = IPSET_ATTR_CMD_MAX,
2201 .policy = ip_set_setname_policy,
2204 .call = ip_set_uadd,
2205 .attr_count = IPSET_ATTR_CMD_MAX,
2206 .policy = ip_set_adt_policy,
2209 .call = ip_set_udel,
2210 .attr_count = IPSET_ATTR_CMD_MAX,
2211 .policy = ip_set_adt_policy,
2213 [IPSET_CMD_TEST] = {
2214 .call = ip_set_utest,
2215 .attr_count = IPSET_ATTR_CMD_MAX,
2216 .policy = ip_set_adt_policy,
2218 [IPSET_CMD_HEADER] = {
2219 .call = ip_set_header,
2220 .attr_count = IPSET_ATTR_CMD_MAX,
2221 .policy = ip_set_setname_policy,
2223 [IPSET_CMD_TYPE] = {
2224 .call = ip_set_type,
2225 .attr_count = IPSET_ATTR_CMD_MAX,
2226 .policy = ip_set_type_policy,
2228 [IPSET_CMD_PROTOCOL] = {
2229 .call = ip_set_protocol,
2230 .attr_count = IPSET_ATTR_CMD_MAX,
2231 .policy = ip_set_protocol_policy,
2233 [IPSET_CMD_GET_BYNAME] = {
2234 .call = ip_set_byname,
2235 .attr_count = IPSET_ATTR_CMD_MAX,
2236 .policy = ip_set_setname_policy,
2238 [IPSET_CMD_GET_BYINDEX] = {
2239 .call = ip_set_byindex,
2240 .attr_count = IPSET_ATTR_CMD_MAX,
2241 .policy = ip_set_index_policy,
2245 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
2247 .subsys_id = NFNL_SUBSYS_IPSET,
2248 .cb_count = IPSET_MSG_MAX,
2249 .cb = ip_set_netlink_subsys_cb,
2252 /* Interface to iptables/ip6tables */
2255 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
2259 int copylen = *len, ret = 0;
2260 struct net *net = sock_net(sk);
2261 struct ip_set_net *inst = ip_set_pernet(net);
2263 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2265 if (optval != SO_IP_SET)
2267 if (*len < sizeof(unsigned int))
2270 data = vmalloc(*len);
2273 if (copy_from_user(data, user, *len) != 0) {
2279 if (*op < IP_SET_OP_VERSION) {
2280 /* Check the version at the beginning of operations */
2281 struct ip_set_req_version *req_version = data;
2283 if (*len < sizeof(struct ip_set_req_version)) {
2288 if (req_version->version < IPSET_PROTOCOL_MIN) {
2295 case IP_SET_OP_VERSION: {
2296 struct ip_set_req_version *req_version = data;
2298 if (*len != sizeof(struct ip_set_req_version)) {
2303 req_version->version = IPSET_PROTOCOL;
2304 if (copy_to_user(user, req_version,
2305 sizeof(struct ip_set_req_version)))
2309 case IP_SET_OP_GET_BYNAME: {
2310 struct ip_set_req_get_set *req_get = data;
2313 if (*len != sizeof(struct ip_set_req_get_set)) {
2317 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2318 nfnl_lock(NFNL_SUBSYS_IPSET);
2319 find_set_and_id(inst, req_get->set.name, &id);
2320 req_get->set.index = id;
2321 nfnl_unlock(NFNL_SUBSYS_IPSET);
2324 case IP_SET_OP_GET_FNAME: {
2325 struct ip_set_req_get_set_family *req_get = data;
2328 if (*len != sizeof(struct ip_set_req_get_set_family)) {
2332 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2333 nfnl_lock(NFNL_SUBSYS_IPSET);
2334 find_set_and_id(inst, req_get->set.name, &id);
2335 req_get->set.index = id;
2336 if (id != IPSET_INVALID_ID)
2337 req_get->family = ip_set(inst, id)->family;
2338 nfnl_unlock(NFNL_SUBSYS_IPSET);
2341 case IP_SET_OP_GET_BYINDEX: {
2342 struct ip_set_req_get_set *req_get = data;
2345 if (*len != sizeof(struct ip_set_req_get_set) ||
2346 req_get->set.index >= inst->ip_set_max) {
2350 nfnl_lock(NFNL_SUBSYS_IPSET);
2351 set = ip_set(inst, req_get->set.index);
2352 ret = strscpy(req_get->set.name, set ? set->name : "",
2354 nfnl_unlock(NFNL_SUBSYS_IPSET);
2362 } /* end of switch(op) */
2365 if (copy_to_user(user, data, copylen))
2375 static struct nf_sockopt_ops so_set __read_mostly = {
2377 .get_optmin = SO_IP_SET,
2378 .get_optmax = SO_IP_SET + 1,
2379 .get = ip_set_sockfn_get,
2380 .owner = THIS_MODULE,
2383 static int __net_init
2384 ip_set_net_init(struct net *net)
2386 struct ip_set_net *inst = ip_set_pernet(net);
2387 struct ip_set **list;
2389 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2390 if (inst->ip_set_max >= IPSET_INVALID_ID)
2391 inst->ip_set_max = IPSET_INVALID_ID - 1;
2393 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2396 inst->is_deleted = false;
2397 inst->is_destroyed = false;
2398 rcu_assign_pointer(inst->ip_set_list, list);
2402 static void __net_exit
2403 ip_set_net_exit(struct net *net)
2405 struct ip_set_net *inst = ip_set_pernet(net);
2407 struct ip_set *set = NULL;
2410 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2412 nfnl_lock(NFNL_SUBSYS_IPSET);
2413 for (i = 0; i < inst->ip_set_max; i++) {
2414 set = ip_set(inst, i);
2416 ip_set(inst, i) = NULL;
2417 set->variant->cancel_gc(set);
2418 ip_set_destroy_set(set);
2421 nfnl_unlock(NFNL_SUBSYS_IPSET);
2422 kvfree(rcu_dereference_protected(inst->ip_set_list, 1));
2425 static struct pernet_operations ip_set_net_ops = {
2426 .init = ip_set_net_init,
2427 .exit = ip_set_net_exit,
2428 .id = &ip_set_net_id,
2429 .size = sizeof(struct ip_set_net),
2435 int ret = register_pernet_subsys(&ip_set_net_ops);
2438 pr_err("ip_set: cannot register pernet_subsys.\n");
2442 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2444 pr_err("ip_set: cannot register with nfnetlink.\n");
2445 unregister_pernet_subsys(&ip_set_net_ops);
2449 ret = nf_register_sockopt(&so_set);
2451 pr_err("SO_SET registry failed: %d\n", ret);
2452 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2453 unregister_pernet_subsys(&ip_set_net_ops);
2463 nf_unregister_sockopt(&so_set);
2464 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2465 unregister_pernet_subsys(&ip_set_net_ops);
2467 /* Wait for call_rcu() in destroy */
2470 pr_debug("these are the famous last words\n");
2473 module_init(ip_set_init);
2474 module_exit(ip_set_fini);
2476 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL));